OpenAI’s Project Lily uses outsourced prompt reviewers to manually inspect real ChatGPT conversations in order to improve response quality, while imperfect privacy filters can still expose sensitive user details and memory summaries. The reviewers focus on style and tone rather than factual accuracy, raising concerns that many users may not realize private conversations can be seen by humans. #OpenAI #ProjectLily #ChatGPT
Keypoints
- Project Lily is a secretive OpenAI initiative that uses outsourced prompt reviewers.
- Reviewers manually read authentic ChatGPT conversations from real users.
- Privacy filters are intended to remove personal data, but they can miss sensitive details.
- Reviewers often see memory summaries that reveal context, interests, and sometimes location.
- Reviewers grade style and relevance, not factual accuracy, which can leave errors uncorrected.
Read More: https://securityonline.info/project-lily-openai-chatgpt-privacy/